Drug discovery beyond the rule-of-five
Ming-Qiang Zhang and Barrie Wilkinson
Although a very useful guideline for orally bioavailable small-
molecule drug design, the rule-of-five (also known as
Lipinskis rule of drug-likeness) has to some extent been
overemphasized. Firstly, only 51% of all FDA-approved small-
molecule drugs are both used orally and comply with the rule-
of-five. This does not even include the increasing number of
biologicals of which several have reached blockbuster status.
Secondly, it does not cover natural product and semisynthetic
natural product drugs, which constitute over one-third of all
marketed small-molecule drugs. A more balanced and
programmatic approach to drug discovery should be more
productive than to rely on an overemphasis of rule-of-five
compliance. Rather it should consider proactively the
development of parenteral drugs in parallel to oral drugs and to
consider the development of therapeutic antibodies in parallel
to small-molecule drugs. These are particularly relevant for
efforts against first-in-class and/or particularly challenging
targets such as proteases and those involving proteinprotein
interactions. In addition, more effort should be invested in
natural product research. Emerging novel technologies such as
synthetic biology (genetic engineering of living organisms to
produce small-molecule therapeutics) may address several
challenging issues of natural product-based drug discovery
including synthetic feasibility and ligand efficiency.
Addresses
Biotica Technology Ltd., Chesterford Research Park, Little
Chesterford nr Saffron Walden, Essex CB10 1XL, United Kingdom
Corresponding author: Zhang, Ming-Qiang (ming.zhang@biotica.com)
Current Opinion in Biotechnology 2007, 18:478488
This review comes from a themed issue on
Chemical biotechnology
Edited by Gregory L. Challis and David A. Hopwood
Available online 26th November 2007
0958-1669/$ see front matter
# 2007 Elsevier Ltd. All rights reserved.
DOI 10.1016/j.copbio.2007.10.005
Introduction
The rule-of-five (also known as Lipinskis rule) for Although oral bioavailability is a very important criterion
drug-likeness [1] was originally proposed in response
to the large number of randomly made compound
libraries, mostly because of synthetic feasibility and
sometimes the single-minded chase for potency, by com-
binatorial chemistry in the early 1990s. It states that poor
absorption or permeability of a compound is more likely
when there are >5 hydrogen-bond donors, the molecular
mass is >500, calculated log P is >5, and the sum of
Current Opinion in Biotechnology 2007, 18:478488
nitrogen and oxygen atoms in a molecular is greater than
10 [1]. Natural products and drugs that are substrates of
biological transporters are exceptions to the rule. It was an
attempt to rationalize compound design (in particular
compound library design) so as not to make too polar,
floppy, and large molecules which have a lower chance of
exhibiting desirable pharmaceutical properties and oral
bioavailability. It has raised the awareness of the import-
ance of ADME (absorption, distribution, metabolism, and
elimination) and physico-chemical properties for the suc-
cess of drug discovery among medicinal chemists, and
helped to front-load ADME (and later ADME/toxicity)
screening in the industrial drug discovery process. It is fair
to say that the rule-of-five has had a major impact on the
daily practice of medicinal chemistry across the pharma-
ceutical industry and served as a very useful guideline for
orally bioavailable small-molecule drug discovery.
However, during the past 10 years of incorporating the
rule-of-five into routine drug discovery practice, there
has been a tendency to overemphasize this simple guide-
line, in particular its association with drug-likeness [2].
The worrying signs for misuse and overemphasis of the
rule-of-five for determining drug-likeness can be seen in
many companies and drug discovery conferences. Some
tales the authors have heard include those of senior
managers in some companies not accepting an otherwise
highly promising development candidate because it did
not fulfill all rule-of-five criteria. Others take pride in the
fact that the average molecular size of the compounds
their company has synthesized is smaller (or has better
compliance with the rule-of-five) than those of their
competitors. This type of overemphasis has the danger-
ous potential to overfilter compounds that may otherwise
be good drug candidates, and inevitably results in
reduced productivity for discovering new drugs.
The major limitations of dogmatic adherence to the rule-
of-five are twofold: it overemphasizes oral bioavailability
and excludes natural products.
Oral bioavailability and drug discovery
productivity
for a successful drug, many of the therapeutics used in
todays medical practice are given parenterally, that is, not
by oral administration. A recent analysis of 1204 US FDA-
approved small-molecule drugs [3] showed that 803 drugs
can be dosed orally, 421 can be administered parenterally,
and 275 can be used as topical agents. Of all 1204 drugs
only 73% (885 drugs) passed the rule-of-five test of
which 70% (619 drugs) are actually used orally, that is,
www.sciencedirect.com

just over half of all FDA-approved small-molecule drugs
are both orally administered and satisfy the rule-of-five.
On the other hand, 20% of all oral drugs fail at least one of
the rule-of-five parameters.
Therefore, overemphasis on orally bioavailable drugs
following the rule-of-five can be counter-productive.
One can argue that penicillin G might not be marketed if
it were to be discovered today because of the emphasis
on oral bioavailability. The first orally bioavailable peni-
cillin (phenoxymethylpenicillin, penicillin V) was dis-
covered >15 years after the discovery of penicillin G.
Although it might not take 15 years to find an orally
bioavailable analog of penicillin G with modern tech-
nologies, the delayed availability of such an important
drug would inevitably result in unnecessary patient
suffering.
The discovery of orally bioavailable drugs is not always
easy and can be extremely challenging. An illustrative
example can be seen through the efforts undertaken to
discover orally bioavailable thrombin inhibitors [4].
Heparin, an indirect inhibitor of thrombin, was originally
discovered in 1916 and has since been one of the most
frequently used anticoagulants for the treatment of
venous thrombosis. Heparin is not a single substance
but a family of sulfated glycosaminoglycans. The
most frequently used low-molecular-weight heparins
(LMWHs) in clinical treatment have molecular weights
ranging from 4000 to 15 000 and are not orally bioavail-
able. They are administered intravenously or subcu-
taneously, which is difficult for long-term treatment.
Therefore, there is a need for orally bioavailable thrombin
inhibitors. More than 80 years after the discovery of
heparin, during which period almost all major pharma-
ceutical companies had had a thrombin inhibitor project,
the first orally bioavailable thrombin inhibitor ximelaga-
tran from AstraZeneca was approved in 2003, but sadly
had to be withdrawn from the market in 2006 because of
liver toxicities [5]. Of course the pharmaceutical industry
will continue its efforts to discover orally bioavailable
thrombin inhibitors despite this major setback.
In the meantime, the discovery of parenteral thrombin
inhibitors has been much more successful. Several drugs,
including the direct thrombin inhibitors lepirudin
(launched 1998), argatroban (launched 2000), and bivalir-
udin (launched 2001), and the indirect thrombin inhibitor
fondaparinux (launched 2002), have been introduced into
clinical practice. Their availability has significantly
improved the treatment of patients undergoing percuta-
neous coronary intervention and also patients who are
contraindicated with heparin. Had the industry only
relied on the discovery of orally bioavailable thrombin
inhibitors, no or little improvement would have been
made in the past 80 years for the treatment of these
patients.
www.sciencedirect.com
Drug discovery beyond Rule-of-Five Zhang and Wilkinson 479
It is therefore more productive when the industry takes a
programmatic approach to drug discovery. That is, when
the discovery of orally bioavailable drugs proves difficult,
for example because of difficult targets such as proteases
or those involving proteinprotein interactions, the
decision to progress initial discovery efforts toward par-
enteral drugs may be more productive. This not only
provides medically needed drugs more quickly but also
helps to validate a potential first-in-class target more
quickly before intensive effort on oral bioavailability is
put in place. For example, during an attempt to develop
anti-HIV vaccines, Jiang et al. discovered that a peptide
derived from the HIV-1 surface glycoprotein gp41 C-
terminal heptad repeat region showed anti-HIV activity
when incubated with human T cells [6]. This discovery
was initially counter-intuitive why evolution should
have produced a viral surface peptide that is inhibitory
to its own replication. Further studies of the fusion
process between HIV-1 and CD4 cells showed that the
peptide and its congeners acted as decoys of their natural
analogs and inhibited the entry of HIV-1 into T cells. One
of these congeners enfuvirtide [7], a 36-amino acid pep-
tide corresponding to positions 643678 in gp41, was later
developed by Trimeris and Roche as the first-in-class
HIV fusion inhibitor and was approved by the US FDA in
2003 only 10 years after the first report on the anti-HIV
activity of a gp41 peptide fragment and despite of the
challenging 106-step total synthesis required for its man-
ufacture. Enfuvirtide is administered by subcutaneous
injection. Very soon after the report of anti-HIV activity
of gp41 peptides, efforts were initiated to discover orally
bioavailable gp41 inhibitors [8], but till date, there has
been little reported success. This is not surprising given
the general difficulties in inhibiting proteinprotein inter-
actions with small molecules and the lack of well-defined
binding pockets on the surface of gp41.
Similar experience has been encountered in the discovery
of other modulators of proteinprotein interactions. For
example, using fragment-based drug design (or SAR by
NMR) [9] Abbott scientists were able to discover a
potent low nanomolar inhibitor ABT-737 of Bcl-2 and
Bcl-xL, antiapoptotic proteins that help cancer cells to
evade programmed cell death and are important targets
for anticancer drug discovery. ABT-737 is not orally
bioavailable and does not comply with the rule-of-five
(MW 803, 13 heteroatoms N, O, and S) [10]. Its clinical
trials are conducted by parenteral administration. A
further Bcl-2 inhibitor, obatoclax (GX15-070, Gemin
X), a derivative of the naturally occurring prodiginines
is also under clinical development by parenteral admin-
istration [11].
This difficulty in finding orally bioavailable inhibitors of
proteinprotein interactions may be because of the large
interaction surface areas demanded of the small-molecule
inhibitors to interrupt such interactions. Recently, Abbott
Current Opinion in Biotechnology 2007, 18:478488
480 Chemical biotechnology
has announced that the company is progressing an orally
bioavailable Bcl-2 inhibitor ABT-263 toward clinical
trials, which again does not comply with the rule-of-five
[10,12].
A major technology advancement made in the past 10
years is the emergence of biologicals which promise to
transform the pharmaceutical industry. Although the
concept of therapeutic monoclonal antibodies (MAbs)
has been around for a much longer period, it is only
within the past 10 years that we have witnessed their
major success as therapeutics. For example, 10 years ago
there were only two MAb drugs on the world market,
whereas now there are 19 FDA-approved MAb thera-
peutics including 6 blockbuster drugs. The market in
2005 was worth >$13 billion worldwide, representing an
astounding growth of 37% from mid-1990s. Biotech pipe-
lines are overflowing with >160 MAb-based drugs [13].
It is not only productive to develop MAbs as therapeutics
per se but they could also be used to lead the development
of small-molecule drugs, for example to generate clinical
proof-of-concept for a potentially first-in-class target
because of their high specificity against biological targets,
coupled with the relatively straightforward nature of their
manufacturing. This is probably best illustrated by the
development of tyrosine kinase inhibitors for the treat-
ment of cancer [14]. Earlier encouraging data generated
with MAbs such as cetuximab and trastuzumab has been
beneficial to the development of small-molecule inhibi-
tors such as gefitinib, erlotinib, and lapatinib.
It therefore appears sensible to adopt a more program-
matic approach in discovering new drugs instead of
wholly concentrating on oral bioavailability. That is:
 to progress initially a parenteral candidate before too
much effort and resources have to be invested for an
oral candidate; and
 to progress initially a therapeutic antibody wherever
possible (most probably by parenteral administration)
before too much effort and resources have to be
invested for a small-molecule candidate.
These may be particularly relevant for first-in-class drug
discovery efforts against difficult targets such as proteases
and those involving proteinprotein interactions.
Natural products and drug discovery
productivity
A major class of drug molecules that are excluded from the
original analysis which generated the rule-of-five are
natural products. This is a serious defect as we know that
a large percentage of marketed drugs are natural products
and their semisynthetic derivatives. A recent analysis of all
drugs approved worldwide between 1981 and 2006 showed
that 34% of all small-molecule drugs are natural products or
Current Opinion in Biotechnology 2007, 18:478488
their direct semisynthetic derivatives [15]. The impact of
natural products is even more profound among drugs that
are used to treat severe and/or life-threatening diseases
such as cancer and infectious diseases. Among all approved
small-molecule anticancer drugs (total 155), 47% are either
natural products or direct semisynthetic derivatives of
them. Similar dominance of natural products is observed
among the anti-infectives, and amazingly >75% of
approved antibacterials are natural products or their semi-
synthetic derivatives (74 out of 98). It is impossible to
imagine where human health would be without these
natural products! Furthermore, during the period 2000
2006, when high-throughput synthesis should have had
sufficient time in operation to produce approved drugs,
approximately 50% of the approved small-molecule drugs
were still those related to natural products [15].
The high productivity of natural product-based drug
discovery may be related to the fact that their chemical
structures have been biologically prevalidated by evol-
utionary selection which defines structural prerequisites
for binding to proteins. For example, an analysis of over
154 000 natural products showed that the majority of
them have molecular volumes ranging from 100 to
500 A3 [16]. In the meantime, the volumes found in
over 18 000 binding cavities of protein targets range from
300 to 800 A3. Therefore, the average volumes of natural
products correlate with the average dimensions of protein
cavities (note that protein ligands often do not fill the
entire volume of a given protein cavity). Although many
natural products from plants and microorganisms are not
meant to bind to human proteins, many human proteins
consist of the same building blocks and contain similar
structural domains to the targets with which natural
products have coevolved.
Natural products therefore represent privileged chemical
starting points for drug discovery, and yet they have quite
distinct structural characteristics from synthetic molecules.
For example, when analyzed by either a size-independent
chemistry space filter or support-vector-machine
approach, natural products exhibit better scores of drug-
likeness than synthetic compounds [17]. Further, natural
products contain on average twice as many oxygen atoms
and three times fewer nitrogen atoms than synthetic drug
molecules [17]. They also contain a slightly higher number
of hydrogen-bond donors than do synthetic drugs. Natural
products contain approximately four times more chiral
centers and far fewer aromatic rings, a fact which may
engender upon natural products better selectivity when
binding to stereo-defined sites.
Better exploration of the biologically prevalidated struc-
tural scaffolds of natural products could increase our
success rate of finding more active leads. For example,
the research group of Waldmann, in collaboration
with Novartis, have devised a hierarchical structural
www.sciencedirect.com
 Drug discovery beyond Rule-of-Five Zhang and Wilkinson 481

classification of natural products (SCONP) based on
structural similarities and the number of rings in their
scaffolds [16]. Each node of this hierarchical system
contained a core structure found in a number of natural
products and was annotated with biological sources and
activities wherever they were known. The utility of such a
system is that one can navigate and link a structurally
complex natural product leading to a set of synthetically
more feasible scaffolds, for example those most
frequently found in natural products containing two to
four rings. In combination with protein structure sim-
ilarity clustering (PSSC), which identifies similar ligand-
binding cores of different proteins [18], researchers would
be able to select a structurally less complex natural
product scaffold that could potentially bind to the same
protein target of the complex lead. This structurally less
complex and synthetically more feasible scaffold could
then be used as a novel chemotype or chemical starting
point to build targeted compound libraries. The attrac-
tiveness and difference of this approach to fragment-
based approach is that the less complex scaffolds are
themselves found in natural products (not necessarily
linked to the complex lead by evolution or biosynthesis),
and are therefore also biologically prevalidated.

Table 1

US FDA-approved polyketide drugs (up to December 2006)

Compound Indication Compound class Producing organism Semi-synthetic

Oxytetracycline Anti-bacterial Tetracyclines Actinomycete N
Methacycline Anti-bacterial Tetracyclines Actinomycete Y
Democycline Anti-bacterial Tetracyclines Actinomycete Y
Clomocycline Anti-bacterial Tetracyclines Actinomycete Y
Minocycline Anti-bacterial Tetracyclines Actinomycete Y
Lymecycline Anti-bacterial, Tetracyclines Actinomycete Y
anti-malarial,
anti-protozoal
Tetracycline Anti-bacterial Tetracyclines Actinomycete N
Doxycycline Anti-bacterial, Tetracyclines Actinomycete Y
anti-malarial
Tigecycline Anti-bacterial Tetracyclines Actinomycete Y
Rifabutin Anti-bacterial Ansa-macrolide Actinomycete Y
Rifampin Anti-bacterial Ansa-macrolide Actinomycete Y
Rifaximin Anti-bacterial Ansa-macrolide Actinomycete Y
Clarithromycin Anti-bacterial Macrolide (14 membered) Actinomycete Y
Azithromycin Anti-bacterial Macrolide (14 membered) Actinomycete Y
Telithromycin Anti-bacterial Macrolide (14 membered) Actinomycete Y
Dithromycin Anti-bacterial Macrolide (14 membered) Actinomycete Y
Erythromycin Anti-bacterial Macrolide (14 membered) Actinomycete N
Roxithromycin Anti-bacterial Macrolide (14 membered) Actinomycete Y
Mupirocin Anti-bacterial Mupirocin Pseudomonad N
Amphotericin B Anti-fungal Polyene Actinomycete N
Candicidin Anti-fungal Polyene Actinomycete N
Natamycin Anti-fungal Polyene Actinomycete N
Nystatin Anti-fungal Polyene Actinomycete N
Idarubicin Anti-neoplastic Anthracycline Actinomycete Y
Doxorubicin Anti-neoplastic Anthracycline Actinomycete N
Epirubicin Anti-neoplastic Anthracycline Actinomycete Y
Daunorubicin Anti-neoplastic Anthracycline Actinomycete N
Valrubicin Anti-neoplastic Anthracycline Actinomycete Y
Rapamycin Immunosuppresive, FKBP12-binding Actinomycete N
anti-proliferative
FK506 Immunosuppresive, FKBP12-binding Actinomycete N
anti-proliferative
Pimecrolimus Anti-inflammatory FKBP12-binding Actinomycete Y
Simvastatin Anti-cholesterimic Statin Fungi Y
Pravastatin Anti-cholesterimic Statin Fungi Y
Lovastatin Anti-cholesterimic Statin Fungi N
Ivermectin Anthelmintic, Avermectin derivative Streptomyces Y
anti-nematode,
anti-protazoal
Bleomycin Anti-neoplastic Bleomycin Streptomyces N
Hesperetin Anti-cholesterimic Hesperetin Plant N
Etoposide Anti-neoplastic Etoposide Plant Y

Data was complied through analysis of DrugBank (http://redpoll.pharmacy.ualberta.ca/drugbank/ [38]).

www.sciencedirect.com Current Opinion in Biotechnology 2007, 18:478488

482 Chemical biotechnology

Synthetic biology for natural product drug
discovery
Indeed one of the greatest challenges for natural product-
based drug discovery has been their structural complexity
and low synthetic feasibility. However, recent progress in
organic chemistry [19] and the emergence of novel tech-
nologies such as synthetic biology (biosynthetic engin-
eering) may help to overcome this issue.
their components to address various problems that cannot
be solved using naturally occurring entities [20]. In
relation to drug discovery, the relevance of synthetic
biology can be considered largely synonymous with
metabolic or biosynthetic engineering which engineers
the genetics of living systems (e.g. microorganisms) using
recombinant DNA technology toward desired end pro-
ducts [21].

Synthetic biology is an emerging scientific field which Although the structures of natural products can be highly
seeks to understand and design biological systems and/or complex and diverse, their biosynthesis can be accom-

Figure 1

Polyketide biosynthesis on a modular PKS as exemplified by the biosynthesis of rapamycin. Biosynthesis is initiated by loading of the PKS with
4,5-dihydroxycyclohexe-1-ene carboxylic acid, with reduction by the loading module ER domain, and subsequent rounds of chain extensions with
malonyl-CoA and methylmalonyl-CoA, in which various levels of reductive processing of the newly formed b-keto groups occur. The modular
organization of the PKS is shown as a series of spherical domains: CoL, CoA-ligase like domain; ER, enoylreductase; ACP, acylcarrier protein;
AT, acyltransferase; DH, dehydratase; KR, b-ketoacylreductase; KS, b-ketoacylsynthase. Biosynthesis is completed by the amidation of the chain
with L-pipecolic acid and macrocylization. The resulting prerapamycin is then converted to rapamycin by the action of several post-PKS acting
monooxygenases and methyltransferase.

Current Opinion in Biotechnology 2007, 18:478488 www.sciencedirect.com


plished by a remarkably few simple types of reactions.
One example is the biosynthesis of the aliphatic, or
reduced, polyketides [22], a structurally diverse class of
natural products that include drugs such as lovastatin,
erythromycin, rifamycin, amphotericin, and rapamycin
among others (Table 1). The biosynthesis of their poly-
ketide core structures involves one major type of reac-
tion a decarboxylative Claisen-type condensation in
which various acyl-CoA monomers are selected by a
polyketide synthase and added, one by one, to the grow-
ing polyketide chain which is tethered covalently to the
PKS (Figure 1). The resulting b-keto esters thus formed
can be further (optionally) reduced to a hydroxyl group,
dehydrated to give an olefin, and finally reduced to a
saturated methylene unit. These few simple chemical
transformations occur with specific stereochemical out-
comes, and in combination, generate tremendous struc-
tural diversity. The simplicity of polyketide biosynthesis
is further facilitated by the modular organization of the
multienzyme PKSs in which highly similar modules a
group of enzymatic domains responsible for one ketide
extension act as an assembly production line, with the
structure of the chemical product hardwired into the
DNA sequence encoding the PKS. Whilst exceptions to
the one module for one ketide extension rule do exist
[23], in general this simple template-based process pro-
vides a direct and easily understood relationship between
DNA sequence and chemical structure, and a paradigm
for the rational engineering of such systems to generate
new natural product analogs biosynthetically [22].
As an example, this technology has been successfully
used to generate novel rapamycin analogs potentially
useful as immunosuppressants, anticancer and anti-
inflammatory drugs. Rapamycin (Figure 2) is a polyketide
macrolide originally discovered from the culture of Strep-
tomyces hygroscopicus NRRL5491 collected from Easter
Figure 2
Structures of rapamycin and its semisynthetic analogs in clinical use or trials.
www.sciencedirect.com
Drug discovery beyond Rule-of-Five Zhang and Wilkinson 483
Island (Rapa Nui). It is a very potent inhibitor of the
serine-threonine kinase target of rapamycin (TOR)
involved in the phosphatidylinositol 3-kinase (PI3K)/
Akt (protein kinase B) signaling pathway that mediates
cell survival and proliferation [24]. It inhibits TOR by first
binding to the immunophilin FK506-binding protein 12
(FKBP-12) and the resulting binary complex then binds
to TOR at an allosteric site, not competing directly with
ATP. It is therefore the most selective kinase inhibitor
known till date.
The structural complexity of rapamycin has been a major
challenge for analog preparation, and there have been
several total syntheses reported over the years [25]. None
of these, however, is less than 50 steps, nor offers an
overall yield >0.5%. It is therefore impractical to use any
of these for lead optimization. As a consequence, the
rapamycin analogs that have progressed to market or
clinical development are all simple semisynthetic deriva-
tives at a single structural point, the secondary hydroxyl
functionality at C40 (Figure 2) [24]. The structural sim-
ilarity of these compounds means they share very similar
biological profiles with little significant improvement, for
example, over poor pharmaceutical or pharmacokinetic
properties.
The biosynthesis of rapamycin is catalyzed by a mixed
type I PKS/nonribosomal peptide synthetase (NRPS)
system (Figure 1) [26]. The PKS uses a shikimate-derived
4,5-dihydroxycyclohex-1-enecarboxylic acid starter unit
and carries out a total of 14 successive steps of polyketide
chain extension. The NRPS then incorporates an
L-lysine-derived L-pipecolic acid moiety into the chain,
followed by cleavage from the enzyme to produce the
key intermediate prerapamycin [27]. Prerapamycin is
further modified by methyltransferases (RapI, RapM,
and RapQ) and cytochrome P450 monooxygenases
Current Opinion in Biotechnology 2007, 18:478488
484 Chemical biotechnology
Figure 3
Mutasynthesis of prerapamycin analogs using the genetically engineered strain Streptomyces hygroscopicus MG2-10 and feeding exogenous starter
acid analogs [28].
(RapJ and RapN) to yield the fully processed rapamycin
molecule [28].
Interestingly, when a region of DNA including the genes
thought to encode those post-PKS acting methyltransfer-
ases and monooxygenases (rap KIJMNOQL) were excised,
the resultant mutant S. hygroscopicus MG2-10 was unable
to produce prerapamycin unless fed with exogenous
pseudo-starter acid 3,4-dihydroxycyclohexane-carboxylic
acid or complemented with the gene rapK (Figure 3)
[29]. This indicated that the product of rapK is involved
in starter acid biosynthesis and/or its regulation.
The production of prerapamycin by feeding exogenous
starter acid to the MG2-10 mutant, combined with the
restoration of post-PKS structural changes by expressing
all or selected post-PKS genes (rapI,J,M,N,O,Q), opened
up the possibility of the combinatorial biosynthesis of a
wide range of rapamycin analogs [29]. For example, it
was demonstrated that feeding other carboxylic acids in
place of the starter unit acid led to the production of
various rapamycin analogs. One of these analogs, BC210
(Figure 4), which has the metabolically labile C39-meth-
oxy group of rapamycin removed from its structure has
shown significantly improved metabolic stability (t1/2
59 min when incubated with human liver microsomes)
and Caco-2 permeability (Papp 29 nm/s) compared to
rapamycin (t1/2 40 min with human liver microsomes
and Papp 2 nm/s) [30]. BC210 has low nanomolar inhibi-
Current Opinion in Biotechnology 2007, 18:478488
tory potency against TOR kinase activity and exhibits
potent anticancer activities in various in vitro and in vivo
assays. It is not a substrate of the P-gp efflux pump and
has good oral bioavailability. Of particular interest is the
effective penetration of the bloodbrain barrier by
BC210. When given intravenously (i.v.) 3 mg/kg in mice,
it accumulates in the brain resulting in the brain to blood
area under curve (AUC) ratio of 1.6, a property which is
unique among published rapamycin analogs (cf. the ratio
for rapamycin is 0.25). This pharmacokinetic property of
BC210 makes the compound a highly promising candi-
date as a potential therapeutic for the treatment of
brain tumors such as glioblastoma multiformes, or
neurodegenerative diseases, either alone or in combi-
nation with other therapeutic agents [31].
The utility of such genetically engineered biosynthesis
approaches for drug discovery has additional benefit
compared to the more widely used semisynthetic
approaches. For example, semisynthetic derivatization
often produces analogs with larger molecule sizes than
the natural product lead, leading to reduced ligand effi-
ciency [32]. Biosynthetic modifications need not suffer
from this drawback. The 39-desmethoxy rapamycin
BC210 (Figure 4) has similar FKBP-12/TOR binding
affinity to rapamycin but its molecular weight is slightly
less than rapamycin. On the contrary, all semisynthetic
derivatives of rapamycin currently marketed or under
clinical development (Figure 2) have larger molecular
www.sciencedirect.com

Figure 4
Design and biosynthetic preparation of BC210: removal of the methoxy moiety at C39 of rapamycin using mutasynthesis provides a compound
with enhanced metabolic stability versus human liver microsomes [29].
weight than rapamycin and equal or less affinity to the
targets FKBP-12/TOR.
This advantage of being able to maintain or increase
ligand efficiency by biosynthesis-based lead optimization
has also been observed in several other cases. For
example, optimization of the naturally occurring heat-
shock protein-90 (Hsp90) inhibitor macbecin I by a
synthetic biology approach has generated a nonquinone
derivative that has significantly increased affinity to
Hsp90 (Kd 0.007 mM) and a reduced molecular weight
(MW 519) compared to the lead molecule (Kd 0.24 mM,
MW 559 for macbecin I) (Figure 5) [33]. On the contrary,
semisynthetic optimization of another naturally occurring
Hsp90 inhibitor geldanamycin has resulted in a 17-ally-
lamino derivative, tanespimycin, that has similar affinity
to Hsp90 (Kd 1.3 mM) but increased molecular weight
(MW 586) compared to the lead geldanamycin (Kd
www.sciencedirect.com
Drug discovery beyond Rule-of-Five Zhang and Wilkinson 485
1.2 mM, MW 561) (Figure 5) [34]. A further advantage
of the nonquinone derivative produced by biosynthetic
engineering over the semisynthetic tanespimycin is that it
does not undergo redox cycling and should have signifi-
cantly reduced off-target toxicity and consequently an
increased therapeutic window.
An increased therapeutic window has also been
achieved with a series of borrelidin analogs without
increasing their molecular sizes [35]. The biosynthetic
derivative BC194 exhibited decreased toxicity but
increased antiangiogenic activity against human umbi-
lical vein endothelial cells (HUVEC) compared to bor-
relidin, and its reduced toxicity was confirmed in an in
vivo study in nude mice. The maximum tolerated dose
(MTD) of BC194 was >130 mg/kg, i.v., whereas the
MTD of borrelidin was found to be in the range of
515 mg/kg, i.v. [35].
Current Opinion in Biotechnology 2007, 18:478488
486 Chemical biotechnology
Figure 5
Comparison of semisynthetic and synthetic biology-based lead optimization approaches for natural products, exemplified by optimization of
ansamycin-based Hsp90 inhibitors. Binding efficiency indices (BEI) = pKd/MW (kDa). For a small-molecule inhibitor of MW 500 and Kd of
1.0 nM, its BEI = 18 [31].
Conclusions
Currently, the majority of industrial efforts in drug dis-
covery, especially in large pharmaceutical companies, are
being invested in discovering small molecule, orally bioa-
vailable drugs that comply with the rule-of-five. How-
ever, only half of all FDA-approved small-molecule drugs
are both used orally and compliant with the rule-of-five
[3]. In other words, nearly half of all small-molecule drugs
are either not used for oral administration or do not comply
with the rule-of-five. This incompatibility between huge
industrial effort and the lower percentage of relevant drugs
in clinical use has almost certainly something to do with
economic considerations oral drugs are more likely to be
blockbusters. Hopefully this is changing with the emer-
gence of blockbuster biologicals and target-based antic-
ancer drugs [36], and the drive to develop drugs for niche
indications [37].
It seems to us that in order to increase drug discovery
productivity and to address seriously un-met medical
needs a programmatic approach to bioavailability could
be adopted. That is first, to consider the development of a
Current Opinion in Biotechnology 2007, 18:478488
parenteral drug candidate in parallel to an oral drug
candidate, in some cases before the availability of an oral
drug candidate and second, to consider the development
of a therapeutic antibody wherever possible in parallel to
a small-molecule drug candidate. These are particularly
relevant for effort against first-in-class drug targets and/
or particularly challenging targets such as proteases and
those involving proteinprotein interactions, where, to
discover an orally bioavailable drug candidate can be
more problematic and slow.
In addition, more effort should be invested in natural
product research. Novel technologies such as synthetic
biology may address many of the challenging issues of
natural product-based drug discovery, for example struc-
tural complexity and synthetic feasibility. Synthetic
biology approaches can in practice be simple and straight-
forward in analog preparation. Genetic alterations can be
made in such a way as to biosynthesize analogs with
predesigned structural modifications, in very much the
same way as achieved by conventional medicinal chem-
istry. Unlike the total synthesis of complex natural pro-
www.sciencedirect.com

ducts, however, it is much more scalable and environ-
mentally friendly for commercial production. It has also
the advantage over traditional semisynthetic approaches
in that it can optimize a structurally complex natural
product lead without the concomitant increase in mol-
ecular size. Therefore, synthetic biology may represent a
better lead optimization approach in terms of maintaining
or increasing ligand efficiency. It can be expected that
with the increasing appreciation of utilities of synthetic
biology in drug discovery, drugs discovered using this
approach will be marketed in the near future.
Acknowledgement
We would like to thank our colleagues at Biotica for their contribution to the
work cited in this review.
References and recommended reading
Papers of particular interest, published within the annual period of
review, have been highlighted as:
 of special interest
 of outstanding interest
1. Lipinski CA, Lombardo F, Dominy BW, Feeney PJ: Experimental
 and computational approaches to estimate solubility and
permeability in drug discovery and development settings. Adv
Drug Deliv Rev 1997, 23:3-25;
Lipinski CA, Lombardo F, Dominy BW, Feeney PJ: Experimental
and computational approaches to estimate solubility and
permeability in drug discovery and development settings. Adv
Drug Deliv Rev 2001, 46:3-26.
A seminal paper in which it was proposed that poor absorption or
permeability of a compound is more likely when there are >5 hydro-
gen-bond donors, the molecular mass is >500, clog P is >5, and the sum
of nitrogen and oxygen atoms in a molecule is greater than 10. Natural
products and drugs that are substrates of biological transporters are
exceptions to the rule.
2. Kubinyi H: Drug research: myths, hype and reality. Nat Rev Drug
Discov 2003, 2:665-668.
3. Overington JP, Al-Lazikani B, Hopkins AL: How many drug
targets are there? Nat Rev Drug Discov 2006, 5:993-996.
4. Quan ML, Smallheer JM: The race to an orally active Factor Xa
inhibitor: recent advances. Curr Opin Drug Discov Dev 2004,
7:460-469;
Schwienhorst A: Direct thrombin inhibitors a survey of
recent developments. Cell Mol Life Sci 2006, 63:2773-2791.
5. Frantz S: Pharma faces major challenges after a year of failures
and heated battles. Nat Rev Drug Discov 2007, 6:5-7.
6. Jiang S, Lin K, Strick N, Neurath AR: HIV-1 inhibition by a
peptide. Nature 1993, 365:113.
7. Poveda E, Briz V, Soriano V: Enfuvirtide, the first fusion inhibitor
to treat HIV infection. AIDS Rev 2005, 7:139-147.
8. Debnath AK: Progress in identifying peptides and small-
molecule inhibitors targeted to gp41 of HIV-1. Expert Opin
Investig Drugs 2006, 15:465-478.
9. Hajduk PJ, Greer J: A decade of fragment-based drug design:
 strategic advances and lessons learned. Nat Rev Drug Discov
2007, 6:211-219.
This paper reviews the development and utilities of this emerging tech-
nology with concrete examples on how it is contributing to drug discovery
in the pharma/biotech industry. The combination of broader sampling of
chemical space and increased hit rates compared to early combinatorial
chemistry and high-throughput screening makes it a powerful tool for hit/
lead identification.
10. Bruncko M, Oost TK, Belli BA, Ding H, Joseph MK, Kunzer A,
Martineau D, McClellan WJ, Mitten M, Ng SH et al.: Studies
leading to potent, dual inhibitors of Bcl-2 and Bcl-xL. J Med
Chem 2007, 50:641-662.
www.sciencedirect.com
Drug discovery beyond Rule-of-Five Zhang and Wilkinson 487
11. Fonseca R, Stewart AK: Targeted therapeutics for multiple
myeloma: the arrival of a risk-stratified approach. Mol Cancer
Ther 2007, 6:802-810.
12. Zhang H, Chen J, Chen Z, Jin S, Nimmer P, Tahir S, Smith M,
Xiao Y, Rosenberg SH, Elmore SW, Tse C: The Bcl-2 family
protein inhibitor, ABT-263, broadly potentiates the cytotoxicity
of other therapeutic agents in vitro. In American Association for
Cancer Research Annual Meeting Proceedings: 2007 April 1418;
Philadelphia (PA) [abstract nr 726]. 2007.
13. Pavlou AK, Belsey MJ: The therapeutic antibodies market to
2008. Eur J Pharm Biopharm 2005, 59:389-396.
14. Baselga J: Targeting tyrosine kinases in cancer: the second
wave. Science 2006, 312:1175-1178.
15. Newman DJ, Cragg GM: Natural products as sources of new
 drugs over the last 25 years. J Nat Prod 2007, 70:461-477.
This latest review together with the previous two by the same authors
(cited in the paper) presents a comprehensive analysis on the impact of
natural products to drug discovery. This includes not only the naturally
occurring compounds and their direct semisynthetic derivatives but also
synthetic drugs that are mimetics of natural products or inspired by
natural products.
16. Koch MA, Schuffenhauer A, Scheck M, Wetzel S, Casaulta M,
 Odermatt A, Ertl P, Waldmann H: Charting biologically
relevant chemical space: a structural classification of natural
products (SCONP). Proc Natl Acad Sci U S A 2005,
102:17272-17277.
An interesting example of how synthetic chemistry can be employed to
biologically relevant or validated structural scaffolds the so-called
biology oriented synthesis (BIOS).
17. Grabowski K, Schneider G: Properties and architecture of
drugs and natural products revisited. Curr Chem Biol 2007,
1:115-127.
18. Charette BD, MacDonald RG, Wetzel S, Berkowitz DB,
Waldmann H: Protein structure similarity clustering: dynamic
treatment of PDB structures facilitates clustering. Angew
Chem Int Ed 2006, 45:7766-7770.
19. Paterson I, Anderson EA: The renaissance of natural products
as drug candidates. Science 2005, 310:451-453.
20. Benner SA, Sismour AM: Synthetic biology. Nat Rev Genet 2005,
6:533-543.
21. Van Lanen SG, Shen B: Progress in combinatorial biosynthesis
for drug discovery. Drug Discov Today Technol 2006, 3:285-292.
22. Staunton J, Weissman KJ: Polyketide biosynthesis: a
millennium review. Nat Prod Rep 2001, 18:380-416.
23. Moss SJ, Martin CJ, Wilkinson B: Loss of co-linearity by modular
polyketide synthases: a mechanism for the evolution of
structural diversity. Nat Prod Rep 2004, 21:575-593.
24. Rubio-Viqueira B, Hidalgo M: Targeting mTOR for cancer
treatment. Curr Opin Investig Drugs 2006, 7:501-512.
25. Nakata T: Total synthesis of macrolides. In Macrolide
Antibiotics: Chemistry, Biology and Practice. Edited by Omura S.
Academic Press; 2002:220-232.
26. Schwecke T, Aparicio JF, Molnar I, Konig A, Khaw LE, Haydock SF,
Oliynyk M, Caffrey P, Cortes J, Lester JB et al.: The biosynthetic
gene cluster for the polyketide immunosuppressant rapamycin.
Proc Natl Acad Sci U S A 1995, 92:7839-7843.
27. Gregory MA, Gaisser S, Lill RE, Hong H, Sheridan RM, Wilkinson B,
Petkovic H, Weston AJ, Carletti I, Lee HL et al.: Isolation and
characterization of pre-rapamycin, the first macrocyclic
intermediate in the biosynthesis of the immunosuppressant
rapamycin by S. hygroscopicus. Angew Chem Int Ed Engl 2004,
43:2551-2553.
28. Gregory MA, Hong H, Lill RE, Gaisser S, Petkovic H, Low L,
Sheehan LS, Carletti I, Ready SJ, Ward MJ et al.: Rapamycin
biosynthesis: elucidation of gene product function. Org Biomol
Chem 2006, 4:3565-3568.
29. Gregory MA, Petkovic H, Lill RE, Moss SJ, Wilkinson B, Gaisser S,
 Leadlay PF, Sheridan RM: Mutasynthesis of rapamycin
analogues through the manipulation of a gene governing
Current Opinion in Biotechnology 2007, 18:478488
488 Chemical biotechnology
starter unit biosynthesis. Angew Chem Int Ed Engl 2005,
44:4757-4760.
A good example of how genetically engineered biosynthesis can be used
effectively to make analogs of a structurally complex natural product
lead.
30. Sheridan RM, Gregory MA, Petkovic H, Gaisser S, Lill RE,
Ward MJ, Kaja AL, Beckman CH, Challis IR, Moss SJ et al.: Brain-
selective mTOR inhibitors as potential therapeutics for
glioblastoma multiforme. In American Association for Cancer
Research Annual Meeting Proceedings: 2006 April 15;
Washington, DC [abstract nr 5959]. 2006.
31. Zhang MQ, Sheridan RM: 39-Desmethoxyrapamycin,
compositions and methods of use thereof. US 7,183,289 B2;
2006.
32. Abad-Zapatero C, Metz JT: Ligand efficiency indices as
guideposts for drug discovery. Drug Discov Today 2005,
10:464-469.
33. Martin CJ, Gaisser S, Vousden WA, Sheehan LJ, Foster T,
Sheridan RM, Nur-E-alam M, Coates N, Moss SJ, Beckmann C
et al.: Towards best-in-class Hsp90 inhibitors: design,
biosynthesis and pre-clinical profiles. In American Association
Current Opinion in Biotechnology 2007, 18:478488
for Cancer Research Annual Meeting Proceedings: 2007 April 14
18; Philadelphia (PA) [abstract nr 5724]. 2007.
34. Tian ZQ, Liu Y, Zhang D, Wang Z, Dong SD, Carreras CW, Zhou Y,
Rastelli G, Santi DV, Myles DC: Synthesis and biological
activities of novel 17-aminogeldanamycin derivatives. Bioorg
Med Chem 2004, 12:5317-5329.
35. Wilkinson B, Gregory MA, Moss SJ, Carletti I, Sheridan RM, Kaja A,
Ward M, Olano C, Mendez C, Salas JA et al.: Separation of anti-
angiogenic and cytotoxic activities of borrelidin by
modification at the C17 side chain. Bioorg Med Chem Lett 2006,
16:5814-5817.
36. Reichert JM, Valge-Archer VE: Development trends for
monoclonal antibody cancer therapeutics. Nat Rev Drug Discov
2007, 6:349-356.
37. Owens J: 2006 drug approvals: finding the niche. Nat Rev Drug
Discov 2007, 6:99-101.
38. Wishart DS, Knox C, Guo AC, Shrivastava S, Hassanali M,
Stothard P, Chang Z, Woolsey J: DrugBank: a comprehensive
resource for in silico drug discovery and exploration. Nucleic
Acids Res 2006, 34:D668-D672.
www.sciencedirect.com